[plt-scheme] Match bug

Bruce Hauman wrote:
>   For list-related administrative tasks:
>   http://list.cs.brown.edu/mailman/listinfo/plt-scheme
> 
> I replied to this earlier,  but my mailer has a bug and the message was
> not posted to the list.  I think I lost several postings that way.

Thanks Bruce.  I got your earlier message, but sent from Matthew.  I 
responded to him, so let me just cut and paste my message here.

>>And I was wondering if the grammar for match could be slightly extended.
>>
>>I believe that (and pat1 ... patn) could be (and lvp1 ... lvpn),
>>likewise for `or' and `not'.
>>
>>This would allow patterns such as
>>
>>   (or ('p x) ...
>>       ('q x) ...)
> 
> 
> Since you are only matching against one object here I am
> suprised that you would want to keep matching a pattern to it
> continually?
> 
> Am I missing something?
> 
> Have a good one,
> Bruce

I'm not sure I follow this question, so let me try to motivate my request...

If we were dealing with xml S-exprs, the pattern (or ('p x) ... ('q x) 
...) would match a homogenous list of p or q elements binding the list 
of contents to x.

Eg.

(match '((p "1") (p "2") (p "3"))
   ((or ('p x) ... ('q x) ...)
    x)
   (_ #f))

Would give ("1" "2" "3").  On the input '((q "8") (q "9")) it would give 
("8" "9").  But for '((p "1") (q "9")) it would give #f, since the list 
doesn't mind it's p's and q's.

This came up for me recently trying to write a structural checker for a 
newsML document (an older draft of newsML).  The DTD has something like 
this:

<!ELEMENT newsitem      (title+,
                         (newsitempart+ | newsobject | text )?,
                         metadata?,
                         handling*,
                         sourcedata?)>

I wanted to write:

(match-lambda
   (('newsitem
     ('title _) ..1
     (or
       ('newsitempart _) ..1
       ('newsobject _)
       ('text _))
     <...>
     )
    #t))

It seems like the natural encoding, but this or pattern is not allowed 
since ('newsitempart _) ..1 is not a pat but an lvp.

As far as I can see the above pattern can't be expressed in the current 
grammar without nesting matches or other funkiness.  I'd be happy to 
hear I'm wrong.  Also, changing the `and' `or' and `not' grammar 
doesn't, as far as I can see, introduce any ambiguity.

-d